// the find
jeecgboot/JeecgBoot
AI 低代码平台「低代码 + 零代码」双驱动！低代码可一键生成前后端代码;零代码可 5 分钟搭建系统;AI Skills 一句话画流程、设计表单、生成整套系统。内置 AI聊天、知识库、流程编排、MCP插件等，兼容主流大模型。引领「AI 生成 → 在线配置 → 代码生成 → 手工合并->AI修改」开发模式，消除 Java 项目 80% 的重复工作，提效而不失灵活。
JeecgBoot is a Java low-code platform built on Spring Boot 3 + Vue 3 that generates CRUD code from database tables and lets non-developers configure forms/workflows visually. It targets Chinese enterprise MIS/OA/ERP development, and has recently bolted on an AI layer (RAG, flow orchestration, MCP, natural language table/form generation) positioned as a Dify alternative. At 47k stars it has a large, active community, primarily in China.
The monolith-to-microservices toggle is genuinely useful — you can develop as a single Spring Boot app and switch to Spring Cloud Alibaba without rewriting code, which covers a real adoption risk. Row/column/field-level data permissions baked into the base controller is something most teams spend weeks building from scratch. The code generator produces runnable front+back code with import/export already wired in, not just skeletons. Multi-database support (MySQL, PostgreSQL, Oracle, DM, KingBase) is real and tested, not just a feature checkbox — important for Chinese government/state-owned enterprise procurement.
The AI features are shallow integrations — it calls external LLM APIs and wraps them in UI, it does not do anything architecturally novel. 'One sentence generates a whole system' is a demo, not production reality. The default password is admin/123456 in the README with no warning about changing it before deployment, which is how these platforms end up on Shodan. The codebase is heavily annotation-driven (MyBatis-Plus + dynamic SQL generation + online forms) which means runtime SQL construction is everywhere — the SqlInjectionUtil filter is a blocklist, not parameterized queries, so injection surface is broad. Documentation and community are almost entirely in Chinese, so non-Chinese teams will hit friction on anything beyond the happy path.